I can confirm the 'data' event in the child process does fire for the incoming packets, but the interval timer's callback never gets executed. I tried this both in a Win7 VM and a Win7 netbook I have. I tried both x86 and x64 binaries on both setups and got the same problem. I should note though that setInterval does work, just not with this benchmark (too much data coming in too fast?).
I do get output however on XP (XP64 to be precise).
On a related note, the throughput benchmark reveals pretty bad results on XP64 (x86 and x64 node binaries both): ~1.84 gbit/sec
This is on an Intel Q9550 with 8GB RAM. On my i7-860 8GB RAM machine running Ubuntu 12.04 x64 (x64 node binary) I get ~14.3 gbit/sec.
Interesting, it works fine with my two win7 setups :-)
And yeah, XP is slow. Who cares?
I didn't realize Windows was that slow, but either problem doesn't concern me that much since I deploy to *nix. However I thought it might be worth posting an issue about because I thought it was interesting that the interval's callback wasn't being executed, but commenting out the relevant net code in the child script shows the interval working just fine.
In my experience XP is always much slower than Windows 7 and later. There's not really anything we can do about that.
And, yes, this is an issue worth investigating. Unfortunately I can't reproduce it (tested it on my macbook and a another laptop). If you could find the time, maybe you try changing the benchmark a little bit to see if it's really a starvation issue. Inserting an empty for loop between sends in in might make it work for example.